Early post‐operative nausea and vomiting: A retrospective observational study of 2030 patients

The overall risk of post‐operative nausea and vomiting (PONV) after general anaesthesia is reportedly 20%–40%. The first episode of PONV may occur early in the post‐anaesthesia care unit (PACU) or later at the ward or after discharge at home in an ambulatory setting. This study aimed to investigate and describe the risk of early PONV in a PACU, and we hypothesised that patients and perioperative factors were associated with early PONV.

whereas others report it as PONV during the entire PACU stay. [6][7][8][9] The risk of early PONV in prospective trials is reportedly around 20%. 10,11 In a randomised controlled trial with 1180 patients, volatile anaesthetics were reported to have a high risk for early PONV (0-2 hours) as compared with propofol anaesthesia, whereas being a child, early PONV and the use of post-operative opioids were the risks for delayed PONV (2-24 hours). 2 Further, studies reporting early PONV are often controlled trials evaluating interventions and using a specified cohort of patients. [3][4][5]8 At the PACU, patients are extensively monitored, and resources are available to evaluate PONV. Further, many hospitals include PONV at PACU in quality registries, and analysis of PONV can be easily accessible. Around half of patients that experience any PONV have PONV at PACU, [10][11][12] and as early PONV is a risk for delayed PONV, 2 early PONV might be considered an important indicator to identify procedures and factors with high overall PONV risks on an institutional level.
The aims of this retrospective exploratory study were to estimate the risk of early PONV after surgical procedures under GA and identify factors associated with early PONV in a mixed-patient cohort.
We hypothesised that in a large mixed-case cohort, patients and perioperative factors associated with early PONV could be identified.

| MATERIAL S AND ME THODS
The study was a retrospective observational review of patients' medical records and was approved by the Regional Ethical Review Board in Umeå (2018-06-12, 2018/233-3, Chairman Anders Iacobeus). The need for informed consent from the patients was waived by the board.
Inclusion criteria were age ≥18 years, surgical procedures conducted under GA and/or mixed GA + regional anaesthesia in the departments of general surgery, orthopaedic surgery, gynaecology, urology, ophthalmology, ear, nose and throat at Sundsvall's hospital, Sweden, between 1 January 2017 and 30 June 2017. GA was defined as the use of either volatile or intravenous anaesthesia or a combination of these two anaesthetics. Patients who underwent surgery solely with regional anaesthesia or light sedation with propofol were excluded. Further criteria for exclusion were procedures with sensitive patient data (ie, abortions), insufficient data from perioperative medical records, patients with re-operation during the same hospitalisation and inaccessibility of the medical records due to a high level of confidentiality.
The departmental routine for premedication involves a basic analgesic, that is, paracetamol and/or a cyclooxygenase-2 inhibitor (etoricoxib), provided that the patient has no contraindications.
Further, for procedures expected to result in increased pain severity, long-acting opioids (peroral controlled-release oxycodone) could be included in the premedication regimen; alternatively, a regional block (epidural analgesia, intrathecal analgesia, upper or lower extremity nerve blocks) was performed. The choice of anaesthesia was based on the type of procedure and local tradition. There were no major changes in this routine during the study period.
A list of all patients scheduled to undergo procedures during the study period was obtained from the local surgical registry (Orbit ® Version 5, Evry Healthcare System AB). Patients under 18 years of age; patients undergoing sensitive procedures (abortions), not requiring PACU stays, undergoing psychiatric procedures (ie, electroconvulsive therapy) and undergoing procedures outside the operating room (OR); and patients who underwent procedures under local anaesthesia were removed from the list. After selecting the eligible patients, study variables were obtained from the surgical registry, medical records, patient's health declaration form, anaesthesia charts and PACU charts.
Based on the known risk factors for PONV, 13,14 variables collected from the sources were preoperative: age, sex, height, weight, body mass index (BMI), American Society of Anaesthesiologists (ASA) physical status, smoking status, history of PONV or motion sickness, same-day surgery or inpatient surgery and elective or emergency surgery; perioperative: surgical procedure and type of anaesthesia, administered PONV prophylaxis, long-acting opioids during anaesthesia (ie, morphine towards the end of the procedure) and the duration of surgery and anaesthesia; and post-operative: time at the recovery unit, events of nausea/vomiting, pain assessments registered on the PACU chart, administered PONV rescue treatment (antiemetics), administered opioids, registration of PONV and data on pain relief obtained in the surgical registry.
The simplified PONV score (Apfel score) was used to estimate the individual risk for PONV. The score was the sum of the factors of the female gender, history of PONV or motion sickness, non-smoking status and anticipated need for post-operative opioids. 15 Anticipated need for post-operative opioids was specific for each type of surgical procedure and was based on the clinical routine at the department.
The PONV prophylaxis was defined as interventions during anaesthesia to reduce the risk of PONV. The departmental routine for PONV prophylaxis was based on the PONV score and published guidelines 13,15 and consisted of utilising total intravenous anaesthesia (TIVA) when possible, ondansetron (4 mg IV), betamethasone (4-8 mg IV) and droperidol (0.5 mg IV). The selection of prophylactic treatment was decided by the attending anaesthesiologist.
Further, PONV prophylaxis aimed at being administered in relation to the PONV risk. To classify if the prophylaxis given to each individual was adequate, we used the relationship between the simplified PONV score and the number of prophylaxis given. We considered that a patient was given optimal prophylaxis when the number

Editorial comment
Post-operative nausea and vomiting (PONV) can be dependent on a variety of perioperative factors. This retrospective register analysis of a mixed-patient cohort revealed that early PONV occurred in every 10th patient and that suboptimal PONV prophylaxis was strongly associated with this. of prophylaxis given was at least one less than the PONV score. 13 Suboptimal prophylaxis was considered when, for example, a patient with a PONV score of three (factors) was given only one prophylaxis.
If the number of prophylaxis given exceeded the optimal amount, we considered it as more than optimal. Our model is a linear simplification of the recommendations of the third consensus guidelines from 2014, 13 with the difference being that with four risk factors, we considered three prophylactic measures as a minimum, whereas the guidelines recommend a minimum of two.
The clinical routine in the PACU was to administer alfentanil or morphine for post-operative rescue treatment of pain and ondansetron for established PONV. All administered drugs were documented on the PACU chart.

| Definition of outcome variables
Early PONV was defined as any documented event of nausea and/or vomiting OR the administration of PONV treatment at the PACU OR the registration of PONV events in the surgical registry. Further, we limited early PONV to the first four post-operative hours in the PACU.

| Data analysis and statistics
Our sample size was primarily based on the resources available for the review of patient records, and we considered that analysis of For the univariate analysis, continuous variables were dichotomised. Univariate analysis with the chi-square test was used to explore associations with early PONV. The odds ratio (OR) with the 95% CI was calculated.
Further, a multivariable logistic regression analysis was conducted.
Variables entered into the model were those with P-values ≤ .25 in the univariate analysis. Factors known to influence PONV were also included. For variables clearly related to each other (ie, 'maximal NRS >5 at recovery unit' and 'rescue opioids at the recovery unit'), we included only one of the variables to reduce the dilution of the model.
Variables based on other variables (simplified PONV score, number of given prophylaxis and suboptimal prophylaxis) were for the same reasons not included in the primary model. We used a backward elimination approach with P > .1 as the significance level for exclusion of the variables from the model. For the multivariate models, the adjusted odds ratio (aOR) with the corresponding 95% CIs is presented.
Additionally, as suboptimal prophylaxis was a highly significant factor in the univariate analysis, we constructed an alternative multivariate model, based on the primary analysis including this variable, and excluded the variables on which suboptimal prophylaxis was based.

| RE SULTS
Data from 4680 procedures were retrieved from the surgical registry, and a total of 2847 eligible patients were identified and screened for inclusion; among them, 86 had received local anaesthesia/sedation, 557 had received regional anaesthesia, 99 had no medical records, 39 had incomplete data and 36 had undergone a re-operation within the same hospitalisation period. After exclusions, the final cohort included 2030 patients (Data S1).
A high risk of PONV (PONV score >2) was seen in 31% of all patients, and 116 patients (5.7%) received suboptimal PONV prophylaxis. For further characteristics of the cohort, including missing values, see Table 1.

| Post-anaesthesia care unit
The mean time spent at the PACU was 96 ± 98 minutes. There were documented assessments of pain in 89% of the patients, and 15% of the patients had maximal pain intensity ≥5 on the numerical rating scale (NRS) or the visual analogue scale (VAS). Opioids were administered to 662 patients (33%) in the PACU (Table 2).
Administration of PONV treatment was the main source for identifying early PONV (9.0%; Figure 1 and Table 2).   Table 4.
In the alternative multivariate model, when suboptimal prophylaxis was entered as an independent variable, the strongest independent factors associated with early PONV were suboptimal prophylaxis (aOR 4.11), rescue opioids at the recovery unit (aOR 2.55) and BMI >35 kg m −2 (aOR 2.08). Further details and other associated factors are presented in Table 5.

| D ISCUSS I ON
In this mixed-patient cohort, we found that every 10th patient undergoing GA experienced early PONV in the PACU. Several factors associated with early PONV were identified. Suboptimal PONV prophylaxis, in addition to previously acknowledged risk factors for PONV, was associated with early PONV.
The risk of early PONV in this study cohort was in the lower range compared with previous studies, where risks of 10%-45% have been reported. 3  Opioids, both long acting administered during the procedure and those given at the PACU, are known to increase the PONV risk. 17,18 In line with these reports, opioids were found to be among the strongest factors associated with early PONV, in both the univariate and multivariate models even in our study cohort.
Volatile anaesthesia is a known risk factor for early PONV, 2 and by using TIVA, the risk can be reduced. 19,20 However, patients anaesthetised with TIVA did not show a reduced PONV risk compared with patients administered volatile agents. There is an obvious risk of selection bias with our retrospective study design, and if there was an overall higher PONV risk for patients receiving TIVA in our cohort, we might not find any differences. Further, it is possible to compensate for the higher PONV risk of volatile agents by providing more prophylaxis. 21 Therefore, our results may also indicate that PONV prophylaxis was given correctly according

P-value (chisquare test)
No 1368 85 (6) Suboptimal prophylaxis Note: The number of patients is the total number of patients in the subgroup.

TA B L E 3 (Continued)
Variables Adjusted odds ratio 95% CI Note: Variables entered in the logistic regression model: BMI >35 kg m −2 , female sex, age < 50 years, history of motion sickness, history of PONV, smoking status, ASA class, outpatient surgery (ambulatory), major surgery, laparoscopic surgery, time of anaesthesia ≥60 min, rapid sequence induction, anticipated need for post-operative opioids, long-acting opioids administered during anaesthesia, regional block, spinal opioids, total intravenous anaesthesia, ondansetron as PONV prophylaxis and betamethasone as PONV prophylaxis; droperidol as PONV prophylaxis; rescue opioids at the recovery unit. Variables were excluded one by one if P > .10. Variables entered in the logistic regression model: BMI >35 kg m −2 , female sex, age < 50 years, history of motion sickness, history of PONV, smoking status, ASA class, outpatient surgery (ambulatory), major surgery, laparoscopic surgery, time of anaesthesia ≥60 min, rapid sequence induction, anticipated need for post-operative opioids, long-acting opioids administered during anaesthesia, regional block, spinal opioids, total intravenous anaesthesia, ondansetron as PONV prophylaxis and betamethasone as PONV prophylaxis; droperidol as PONV prophylaxis; rescue opioids at the recovery unit. Variables were excluded one by one if P > .10.

TA B L E 4
Multivariate analysis of independent risk factors associated with early PONV (n = 1742) to PONV risk and that TIVA may be considered as a prophylactic intervention.
A longer duration of anaesthesia (>60 minutes) was associated with early PONV. Long exposure to surgery and anaesthetic drugs are known to increase the risk of PONV. 13,22 To reduce the risk of PONV, it is crucial to administer proper PONV prophylaxis in relation to an assessed PONV risk using the Apfel score. 13 With our model stating the minimum amount of PONV prophylaxis in relation to PONV risk, we found that early PONV was associated with suboptimal PONV prophylaxis administered to patients. Other studies indicate that even when PONV risk models are implemented in a clinical setting and accompanied with prophylactic recommendations, the PONV risk is still high due to low adherence to guidelines. 23 We may further speculate that PONV risk would decrease if patients with suboptimal prophylaxis were given adequate prophylaxis; but as there might be other confounding factors involved, our result only indicates that it might be a target for improvements.
Patients with three or four risk factors receiving two prophylactic interventions had a higher risk of early PONV (13-14%), indicating that the recommended routine for PONV prophylaxis may not be enough. 13 Indeed, in the recently published fourth consensus guidelines for the management of PONV, a more liberal administration of PONV prophylaxis in patients with risk factors is recommended. 24 By following these new guidelines, potential reduction of the risk for PONV may occur.
Two thirds of patients were considered to have been given more prophylaxis than standard, and PONV risk was lower among those patients compared with patients with an optimal prophylaxis. This is in accordance with the new recommendations of PONV prophylaxis.
However, as the Apfel score was based on chart reviews, we could have missed factors included in the score leading to a wrong classification of patients and resulting in a lower score.
The majority of our patients (53%) underwent endotracheal intubation during the procedures, which might be considered high for a consecutive flow of patients. However, patients receiving regional anaesthesia only were not included in our study cohort, and 25% of our patients underwent emergency procedures, which may lead to more instances of endotracheal intubations when compared with elective procedures. Note: Variables entered in the logistic regression model: BMI >35 kg m −2 , age < 50 years, ASA class, outpatient surgery (ambulatory), major surgery, laparoscopic surgery, time of anaesthesia ≥60 min, rapid sequence induction, long-acting opioids administered during anaesthesia, regional block, spinal opioids, rescue opioids at the recovery unit and suboptimal PONV prophylaxis. Variables were excluded one by one if P > .10. The variable suboptimal prophylaxis was based on the relation between the simplified PONV score (including the following factors: female gender, non-smoking status, history of motion sickness or previous PONV and the anticipated need for post-operative opioids) and the number of PONV prophylaxis given (betamethasone, ondansetron, droperidol and total intravenous anaesthesia). If the number of PONV prophylaxis was >1 less than the PONV score, the prophylaxis was regarded as suboptimal.
TA B L E 5 Multivariate analysis of independent risk factors associated with early PONV using an alternative model including suboptimal PONV prophylaxis as an independent variable (n = 1734) results. To further explore and confirm factors associated with early PONV, we suggest studies to be prospective with a pre-defined standard in the collection of data. With a retrospective design, many of the variables collected may have low accuracies in a mixed-case cohort. Our study was also limited by the variables included, and there are many other possible confounders, for example, analgesic premedication, which might be associated with early PONV.
The strength of our study was that a majority of all adult patients receiving GA in our hospital were included. However, this study describes the PONV risk at only one hospital, thus limiting the generalisability of the results. Furthermore, our study only described the PONV risk up to the first four post-operative hours. Many patients experience their first episode of PONV after discharge from the PACU, and to report PONV during the whole post-operative course, patients should be observed for several days after discharge. 10,11 We consider early PONV as potentially an important indicator for the overall PONV risk, and if used in, for example, audits, it may be a valuable tool in PONV-reducing strategies. To further explore the role of early PONV, we plan to confirm our findings of the risk estimates with national data from the Swedish perioperative registry (SPOR). Our results also warrant prospective multicentre studies on the relations between early and late PONV and associated factors.
In conclusion, we found that almost every 10th patient under GA experienced early PONV and we are, to our knowledge, the first to identify risk factors in a mixed-case cohort. Suboptimal PONV prophylaxis, in addition to previously acknowledged risk factors for overall PONV, was associated with early PONV. Our results also highlight the importance of a more liberal approach to PONV prophylaxis and of adhering to PONV guidelines. Further, we suggest future prospective studies evaluating early PONV as an indicator for overall PONV.

CO N FLI C T O F I NTE R E S T S
JW has received lecture fees from Abbvie Sweden AB. The other authors have no conflicts of interest to declare.